This paper presents a neural network model based on explicit approach using the interrelationship characteristics of direct, diffuse and global solar radiations. The computational algorithm includes the estimation of global, diffuse and direct components through clear sky conditions. The deviations of these estimates from measurements are considered to be due to random weather phenomena characterized by clearness indices. The clearness indices corresponding to direct, diffuse and global components of solar radiation are then mapped with long term monthly mean hourly data of weather-related parameters such as mean duration of sunshine per hour, relative humidity and total rainfall in the artificial neural network (ANN) analysis. The estimates of ANN model exhibit excellent compatibility with observations with overall root mean square error RMSE (%) and mean biased error MBE (%) for the global radiation as 5.19 and -0.194 respectively. The RMSE values for wet months (July, August, September, October) are relatively higher than those of the dry months (January, February, March, April) due to intensive monsoon in Indian region. Contour maps of ANN model calculations of global radiation as a function of latitude, time of the day and month of the year are presented. The model offers promise of being useable for the prediction of direct, diffuse and global components of solar radiation at an arbitrary location. A comparative study of the estimates of present ANN model with NASA surface meteorology and solar-energy data sets exhibits good compatibility. The global solar radiation on tilted planes has also been investigated using isotropic and anisotropic sky conditions. The results seem to be favouring an isotropic model during the year-round cycle at New Delhi (28.58 degrees N). (C) 2014 Elsevier Ltd. All rights reserved.